P.7, bottom, et seq., here:
http://www.harman.com/EN-US/OurCompany/Technologyleadership/Documents/White Papers/AudioScience.pdf
1. That's not his textbook. That is something from Harman that was probably over-edited by some PR flack and approved by his management, maybe. Parts do seem very reminiscent of the book.
2. No doubt the fault is mine, but I can't find that part about reviewers crapping in their pants and blushing from loss of "esteem." More specific location, please?
3. In perusing the link, sure enough, Toole plays a bit fast-and-loose with the difference between variability and average-tendency. Not the same at all and not a defensible interpretation on his part. Indeed, the reviewers might be right in changing their opinions based on subtle differences each test session... not that I would defend golden-eared types usually. Maybe all those high school kids and audio salespeople are just listening for good bass and nothing else - as Toole himself points out.
How come pink noise was the best test stimulus (whatever pink noise is supposed to sound like realistically, eh), followed by a lot of over-produced stuff.... and far down on the list was really demanding kinds of music like brass quintet?
El Greco fallacy:
Stuart Anstis
Last edited:
I haven't seen much data lately. Mostly theory and opinions which is why I made the post.
Maybe you need to twist a few arms here to get more data shared?
http://www.hauptmikrofon.de/theile/ON_THE_LOCALISATION_english.pdf
[It's put me to sleep twice, so far, but the data is there....
]
[It's put me to sleep twice, so far, but the data is there....
]http://www.hauptmikrofon.de/theile/ON_THE_LOCALISATION_english.pdf
[It's put me to sleep twice, so far, but the data is there....
I'm not sure if I have a contralateral or ipselateral ear. 😀
Last edited:
Because the spectrum doesn't change with time like music does. Much easier to here differences with when comparing multiple sources. No shhhh fhhhh thhhh just even noise with no emphasis in any one band.
Rob🙂
Beg your pardon, Robh3606, but the Toole/Olive studies asked people what speakers they preferred, not just if they could detect a difference.
I really don't know what "no emphasis in any one band" could mean if I don't know what "real" pink noise sounds like.
Have another theory?
No . Go back into Tooles book it's in there. It's certainly in the Test facility articles.
Get a set of flat speakers, get a disk with pink noise on it and have a go.
Looking at an RTA display it's a flat line. Bands as it 1/3 or 1/6 Octave
How do you set-up your system at home by ear??
Rob🙂
If you first establish that people cannot detect a difference, then it's meaningless to ask which they prefer. Preference is only valid if there are discernable differences.
Although I believe it's not a far stretch to believe that there will almost always be a difference between dissimilar speakers, it certainly must be established that the people invited to the comparisons, can detect a difference.
I think that "real" pink noise differs from "fake" pink noise in that all frequencies will have equal energy across the frequency spectrum. If I'm mistaken feel free to point it out.
Robh3606 was stating a fact about pink noise, so I have to ask: What theory did you have in mind?
Best Regards,
TerryO
Pink noise means power is inversely proportional to frequency, white noise is of equal power across frequency.
I think a big part of evaluating speakers is determining (by ear) what their spectral response is. In comparisons, we can hear broad imbalance and we can also hear local trends or resonances. To do this, the fuller the spectrum, the better.
A brass (or string) quartet is tough because it is a sparse spectrum. solo instruments or voice might not even cover the full frequency range. Pop recordings with "lots going on" will give a fuller spectrum and can make the task easier.
Pink noise, even if we don't know what it should sound like, excites all frequencies and makes A-B tests between speakers easiest.
David S.
Here's some text from a couple of posts I made over at PE's Tech Talk forum in a thread about capacitor evaluations.
"IIRC, way back when (in the early '90's), Sterephile conducted listening tests on loudspeakers. The tests were conducted blind to the listeners who were comprised of their writers - guys who spend a fair amount of time listening to speakers and music.
The tests were run at their former digs in Santa Fe, NM. The test track list consisted of pink noise and a number of music selections. Too many in my opinion.
In what I believe was their last report published in the January 1993 issue (the original small format sized one), they reported on tests run on 9 different speakers they catagorized as small & inexpensive. For each speaker, pink noise and 15 music selections were listened to and given a numerical rating score. This happened over a period of days.
I entered all the scores in a spreadsheet and ran some basic stats. Pink noise had the smallest standard deviation of all the tracks listened to. This may be due wholly or, in part, to the fact that it is a rather consistent signal. But I think this is the point others here were trying to make."
"Since my last post I found another Stereophile article published later that year (Sept.) wherein they ran similar test panels on 8 floor standing speakers priced in the $1k-$2k range at that time. Again, pink noise gave the most consistent scores by far.
Now comes the kicker: When the individual score (ave. of the listener's scores) for each speaker on just the pink noise track was correlated with the average score for each speaker over all the tracks listened to, the correlation was very good (r=0.889). That's pretty good for only 16 data points.
What this means is they could have done away with most to all of the other 15 tracks and gotten about the same result listening to the pink noise track alone. This seems ample justification for use of pink noise in listening panels - wheter it's caps or speakers.
Thus, a smaller variance in test scores would reduce the 'noise' or overall scatter to the data and possibly allow better discrimination between cap assessments."
"IIRC, way back when (in the early '90's), Sterephile conducted listening tests on loudspeakers. The tests were conducted blind to the listeners who were comprised of their writers - guys who spend a fair amount of time listening to speakers and music.
The tests were run at their former digs in Santa Fe, NM. The test track list consisted of pink noise and a number of music selections. Too many in my opinion.
In what I believe was their last report published in the January 1993 issue (the original small format sized one), they reported on tests run on 9 different speakers they catagorized as small & inexpensive. For each speaker, pink noise and 15 music selections were listened to and given a numerical rating score. This happened over a period of days.
I entered all the scores in a spreadsheet and ran some basic stats. Pink noise had the smallest standard deviation of all the tracks listened to. This may be due wholly or, in part, to the fact that it is a rather consistent signal. But I think this is the point others here were trying to make."
"Since my last post I found another Stereophile article published later that year (Sept.) wherein they ran similar test panels on 8 floor standing speakers priced in the $1k-$2k range at that time. Again, pink noise gave the most consistent scores by far.
Now comes the kicker: When the individual score (ave. of the listener's scores) for each speaker on just the pink noise track was correlated with the average score for each speaker over all the tracks listened to, the correlation was very good (r=0.889). That's pretty good for only 16 data points.
What this means is they could have done away with most to all of the other 15 tracks and gotten about the same result listening to the pink noise track alone. This seems ample justification for use of pink noise in listening panels - wheter it's caps or speakers.
Thus, a smaller variance in test scores would reduce the 'noise' or overall scatter to the data and possibly allow better discrimination between cap assessments."
Terry,
Thanks, that's very kind. Those were the good ol days. I even miss Doug Purl's hilarious cheap shots.
The folks on the bass list had enough experience to understand that audio reproduction is an illusion, and physically a very artificial construct with the reproduced sound field looking very different than the original physical soundfield. As such they had the wisdom to refrain from absolutist positions based on limited or partial objective data.
Dave
An interesting post, thank you.
But you've made an elementary error of logic: being able to tell things about (small variance) has ZERO to do with validity in ranking them.
This point is too self-evident, once mentioned, to need any kind of example to illustrate. Which more or less applies to the interpretive errors in the previous bunch of posts... like the guy who implies (erroneously) you can have preferences without discrimination but doesn't see you can have discrimination without preferences.
I suppose you can have speaker preferences - even ones consistent across the whole world - on some basis besides "realism" (don't ask me to define that). But wouldn't that be strange and disheartening to audiophiles. "There's something about that pink noise... that just gets to me... love it... maybe it is the peak at 3500 Hz..." For sure, pink noise and over-produced female vocalists do not constitute the sound sample I know about.
Dave S - brass quintets include tubas and the spectrum is quite wide. If I had just one recording to test with, might be the large Lincoln Center Jazz Orchestra playing Marsalis' "The Big Train."
BTW, Toole showed that raters have a real bad time discriminating between an upper bass trough and a low treble boost. Interesting implications there for the whole enterprise of listener testing.
Last edited:
IME, a mono signal in stereo has a different tonal balance than a mono signal in mono. Anyone here can easily try it for themselves with pink noise on their own system, by facing one speaker with the other muted, then facing center between two with both playing at -6 dB.
Some (most?) of the difference is probably due to the difference in the early reverberant field set up by 2 speakers rather than 1. Without an anechoic chamber, its hard to say how much is HRTF related, but the science says its there, per my first post. The variance in HRTF from person to person should also explain some of why we don't all perceive tonal balance in a speaker the same (one of many reasons: training, preference etc etc), because the change in perceived tonal balance varies from user to user over azimuth, but the speakers are fixed.
And yes, a cello sounds tonally like the same cello whether its 30 deg to the right or directly in front of us (because we adapted to our HRTF, or we'd go crazy). But that's a different scenario than two cellos in perfect time but placed at +/- 30 deg.
I also don't buy that engineers eq for this HRTF effect. I've asked for proof that this is commonly be done, none has been offered. Given the wide variety of speakers used in mastering, and the wide variety in environemnts, its moot anyway.
The time delays of the crosstalk are also short enough that the "comb filtering" appears to not be perceived as colouration. From "ON_THE_LOCALISATION_english.pdf":
"The listener should be positioned in front of two loudspeakers that are set up in an anechoic chamber and that radiate coherent white noise (at a suitably low level). Lateral head movements cause lateral displacements of the phantom source (timebased stereophony), whilst the sound colour hardly changes. Moving the head in the same manner whilst covering one ear will lead to clearly perceivable sound colour changes. The sound colouration, which is caused by the comb filter and which can be detected at each ear individually, disappears when listening binaurally, i.e. as soon as a phantom source arises. Its occurrence implies the suppression of sound colouration.
This effect was first pointed out by THEILE 1978."
I fully agree that the world would be great if everyone mastered and listened in the same room with the same basic speaker response. I've appreciated that paradox for 20 years (I think Heyser pointed it out 40 yrs ago). But we have to make do with the real world!
I'm wondering if many here actually read the Toole study on preference from '85? I just read it again as a rfresher a few weeks ago. He wasn't looking at 1 or 2 dB variances and making conclusions. He was showing that big peaks and dips (well, more so peaks) were less preferred.
I've also pointed out in the past that Toole tested in an IEC room, and that he didn't look at how rankings would change with differences in room contribution. He also didn't explicitely look at how non flat on axis may or may not be perceived as more preferable, if counter-balancing errors in the response of the early reflection contribution.
So there's lots of room for plausibility that non-flat can be preferred as more accurate than flat in an appreciable percentage of circumstances.
Dave
Last edited:
I don't agree with this. I have always noticed that a slight shift while nearly centered on a pair of loudspeakers will show comb filtering. It is independent in each ear, meaning that there is a head position where it goes away for the left ear and another for the right. (The comb filtering goes away when the two speaker's path lengths are equal, only one ear can be in that position at a time.) What I primarily hear is treble level coming and going as a broad range of HF falls into a cancelation notch.
I can hear it in a living room. I'm sure it would be more obvious in an anechoic chamber. If the author is implying that the response of one ear can fill in holes in the response of the other, that certainly isn't the case.
I guess the question is what "hardly changes" means.
David S.
DDF, the good folks here will hopefully display the wisdom to refrain from absolutist positions based on limited or partial objective data and even absolutists positions w/o any data at all. Well, lets ignore the last part of that, then we'll have no discussion. A one way discussion ceases to be one--I think that's called a lecture or a speech. That's where this line of reasoning can go so let's just actually refrain from it. There is no use in 2 or more opposing retort immune diatribes. I do genuinely appreciate the high-mindedness of such an elegant theory as yours or whomever first thought of it.
It's really not that mastering engineers are compensating for HRTF--that would be foolish b/c of the innumerable differences in end users among several other things. I know I've said this before, but it's that the HRTF issue(s) is being compensated for (if it actually exists depending on your view of the problem) by mastering the recording. They're going for good sound(hopefully) and/or to create an auditory illusion(often helped by a video cue 😉 ). Thus HRTF is compensated for as the mid/treble end is not a few dBs hot in their opinion on their studio monitors and other speakers they've mastered the mix on. Playback that recording on speakers that radically deviate from previous conditions (like the Orion) and all bets are off.
Yes, 2 identical (fill in whatever here)_______ playing in time, side by side at 30 degrees may not sound as one in the middle for many reasons. Hmm, perhaps we should use a center channel speaker? Maybe that's why I like a center channel...... I think I've read something on this somewhere. 🙂 One big problem with compensating for HRTF in the speaker intended for stereo (suppose there was actual supported evidence for it) is that most people will not stick their head in a vise and/or near field to listen (except the rare audiophile) and why should we short change anyone outside of the sweet spot even if the sweet spot was actually sweetened by the compensation. Again, that's just tip of the iceberg stuff, the depths of which are beyond a web post. When you factor in just the things DDF mentioned, does it seem worth it to even try? Again, that's up to you, but there is definitely a need to approximate a recording in many instances.
We should all read the El Greco Fallacy Ben linked earlier. Makes it seem reasonable/plausible HRTF is not a factor in individual preferences/perceptions in absents of an acute auditory/neurological event. We all grew up with our own HRTF. What else might be deduced from that is up to the reader. Individuals are more effected by what they see than what it seems most audiophiles believe. I'll give a few examples: Anyone heard of the McGurk effect? YouTube - McGurk Effect Ethan Whiner has demonstrated some illusions in this video as well as many other useful insights into more audio related sensory phenomena: YouTube - Audio Myths Workshop
Look at Audio Musings by Sean Olive: The Dishonesty of Sighted Listening Tests specifically for loudspeaker related issues. Of course just as interesting is the rating variation of the same speaker in different positions under blind conditions/lack there of sighted. The implications of that are many.
Regards,
Dan
It's really not that mastering engineers are compensating for HRTF--that would be foolish b/c of the innumerable differences in end users among several other things. I know I've said this before, but it's that the HRTF issue(s) is being compensated for (if it actually exists depending on your view of the problem) by mastering the recording. They're going for good sound(hopefully) and/or to create an auditory illusion(often helped by a video cue 😉 ). Thus HRTF is compensated for as the mid/treble end is not a few dBs hot in their opinion on their studio monitors and other speakers they've mastered the mix on. Playback that recording on speakers that radically deviate from previous conditions (like the Orion) and all bets are off.
Yes, 2 identical (fill in whatever here)_______ playing in time, side by side at 30 degrees may not sound as one in the middle for many reasons. Hmm, perhaps we should use a center channel speaker? Maybe that's why I like a center channel...... I think I've read something on this somewhere. 🙂 One big problem with compensating for HRTF in the speaker intended for stereo (suppose there was actual supported evidence for it) is that most people will not stick their head in a vise and/or near field to listen (except the rare audiophile) and why should we short change anyone outside of the sweet spot even if the sweet spot was actually sweetened by the compensation. Again, that's just tip of the iceberg stuff, the depths of which are beyond a web post. When you factor in just the things DDF mentioned, does it seem worth it to even try? Again, that's up to you, but there is definitely a need to approximate a recording in many instances.
We should all read the El Greco Fallacy Ben linked earlier. Makes it seem reasonable/plausible HRTF is not a factor in individual preferences/perceptions in absents of an acute auditory/neurological event. We all grew up with our own HRTF. What else might be deduced from that is up to the reader. Individuals are more effected by what they see than what it seems most audiophiles believe. I'll give a few examples: Anyone heard of the McGurk effect? YouTube - McGurk Effect Ethan Whiner has demonstrated some illusions in this video as well as many other useful insights into more audio related sensory phenomena: YouTube - Audio Myths Workshop
Look at Audio Musings by Sean Olive: The Dishonesty of Sighted Listening Tests specifically for loudspeaker related issues. Of course just as interesting is the rating variation of the same speaker in different positions under blind conditions/lack there of sighted. The implications of that are many.
Regards,
Dan
Did anyone else watch Seinfeld when it was on?
Sometimes the discussions that go on for the most pages remind me of the whole premise behind Seinfeld 😉
Sometimes I wonder if we put as much passion in to real life issues as we do on these ultra hypothetical topics. Just think what we could really accomplish in the world 😉
Sometimes the discussions that go on for the most pages remind me of the whole premise behind Seinfeld 😉
Sometimes I wonder if we put as much passion in to real life issues as we do on these ultra hypothetical topics. Just think what we could really accomplish in the world 😉
- Status
- Not open for further replies.